AI System Reduces Hospital Deaths by 26%, Transforming Real-Time Patient Care

A groundbreaking AI-driven tool, CHARTwatch, has significantly improved patient outcomes at St. Michael’s Hospital in Toronto, slashing unexpected deaths in its general medicine unit by 26%. Designed to deliver real-time alerts, the machine-learning tool enables clinicians to identify high-risk patients and intervene before their condition worsens. It demonstrates AI’s immense potential in revolutionizing modern healthcare.

In busy hospital settings, quickly detecting signs of patient deterioration is critical. This time gap can mean the difference between life and death for patients whose conditions worsen unexpectedly. CHARTwatch is designed to close this gap, working quietly in the background to analyze over 100 inputs from patients’ medical records, including vital signs, blood pressure, heart rate, and daily lab results. By making dynamic predictions every hour, the AI tool detects subtle changes in a patient’s condition, alerting doctors and nurses when intervention is needed. This cutting-edge system exemplifies the growing role of AI as a “team member” in clinical care, enhancing decision-making without replacing human judgment.

The success of CHARTwatch is grounded in rigorous development, starting in 2017 and undergoing several years of testing before its deployment in October 2020. Researchers analyzed more than 13,000 admissions in St. Michael’s General Internal Medicine (GIM) ward, one of the hospital’s busiest units, caring for patients with some of the most complex medical conditions. Results showed a significant relative reduction of 26% in “non-palliative” deaths—unexpected deaths not associated with end-of-life palliative care. Death rates dropped from 2.1% to 1.6%, a change that translates into meaningful lives saved. Among high-risk patients, non-palliative deaths fell even further, from 10.3% to 7.1%.

In addition to reducing mortality, CHARTwatch helped optimize treatment. Following the tool’s implementation, clinicians provided more proactive care, such as increasing doses of antibiotics and corticosteroids, closely monitoring vital signs, and escalating treatment when necessary. For example, a patient with what initially appeared to be a mild fever was identified by CHARTwatch as being at risk due to an abnormally high white blood cell count—a subtle but critical finding that led to rapid treatment for cellulitis, preventing a life-threatening outcome.

“It’s not replacing the nurse at the bedside; it’s enhancing your nursing care,” said Shirley Bell, Clinical Nurse Educator at St. Michael’s Hospital. By supporting bedside staff and complementing clinicians’ expertise, CHARTwatch ensures earlier detection of deterioration, allowing faster responses that improve survival rates.

The research, published in the Canadian Medical Association Journal, was co-led by Dr. Amol Verma, clinician-scientist at St. Michael’s, and Dr. Muhammad Mamdani, Vice-President of Data Science at Unity Health Toronto. Verma emphasized the tool’s success in real-world clinical care, a rarity among AI technologies. “Very few AI tools have been implemented into clinical settings,” he said. “This is one of the first in Canada that actively helps us care for patients daily.”

While the study’s findings are promising, researchers acknowledge its limitations. Conducted in a single hospital unit during the COVID-19 pandemic, the results may not be generalizable to more extensive or diverse settings. Resource constraints, patient demographics, and systemic challenges could influence CHARTwatch’s outcomes in other hospitals. However, this study provides valuable evidence that machine-learning tools can meaningfully enhance patient care.

AI’s potential in healthcare extends beyond CHARTwatch. Machine learning models are being explored for early cancer detection, hypertension screening, and even concussion diagnosis through brain pattern analysis. In a time when healthcare systems face severe staffing shortages, tools like CHARTwatch offer critical support, supplementing bedside care and enabling more efficient use of medical resources. Dr. John-Jose Nunez, a psychiatrist and researcher from the University of British Columbia, praised the technology as a valuable addition to clinical teams, calling it “one more team member on the clinical care team.”

Looking ahead, the Unity Health team plans to expand CHARTwatch’s use within its network and beyond. Canada’s most prominent hospital data-sharing network, GEMINI, is already working with over 30 hospitals across Ontario to explore AI tools like CHARTwatch in various clinical settings. By leveraging GEMINI’s infrastructure, researchers aim to replicate and refine the technology for broader adoption.

CHARTwatch’s success marks a significant milestone in healthcare AI. Dr. Mamdani noted, “It just sets the groundwork now to deploy these tools well beyond our four walls.” With its ability to save lives through timely intervention, CHARTwatch is a shining example of AI’s transformative potential in healthcare, ensuring patients receive the care they need before it’s too late.